Skip to main content
JetBrains AI Assistant has a built-in OpenAI-compatible endpoints field for custom providers. Tokios gives you a stable, authenticated URL to put there: pair the connector next to your model, register a deployment name, and hand AI Assistant that URL with an sk-tok-… key. No tunnel to babysit, no unauthenticated port on your machine. This works the same way across JetBrains IDEs that ship AI Assistant — IntelliJ IDEA, PyCharm, GoLand, WebStorm, Rider, and others.

​
Prerequisites

  • A running Tokios connector paired with your local model — see Connector install
  • A registered model deployment name — see Register a model
  • A Tokios API key (sk-tok-…) — see API keys
  • A JetBrains IDE with the AI Assistant plugin installed

​
Point AI Assistant at Tokios

1

Open AI Assistant's provider settings

Open Settings | Tools | AI Assistant | Providers & API keys. JetBrains has moved this setting between releases, so confirm the exact menu path and label for your installed AI Assistant version.
2

Find the OpenAI-compatible endpoints section

Look for the OpenAI-compatible endpoints section, used for third-party OpenAI-compatible servers — the exact field labels shown here can vary by version, so check yours.
3

Enter the Tokios URL

Set the endpoint URL to:
4

Paste your Tokios API key

Paste your sk-tok-… key into the API key field for this endpoint. AI Assistant sends it as a Bearer token on every request.
5

Add your deployment as the model

Enter the public deployment name you registered in the Models tab — for example, gemma-tunnel — as the model for this endpoint. This is the value AI Assistant sends in the model field; it is not Ollama’s, llama.cpp’s, vLLM’s, or LM Studio’s local model id.
6

Toggle tool calling if you need it

If AI Assistant exposes a toggle for tool/function calling on this endpoint, enable it. Whether the model actually acts on tool calls depends on the underlying model itself.
Replace gemma-tunnel with the deployment name you registered, and sk-tok-YOUR_KEY with your own key. Never commit either into a shared config file.
AI Assistant’s custom-provider support in JetBrains IDEs is documented as an OpenAI-compatible slot. There is no confirmed custom Anthropic-compatible base URL field, so connect through the OpenAI-compatible endpoint above rather than Tokios’s Anthropic-style endpoint.
A custom endpoint does not power every AI Assistant feature. As of the 2026.1 releases, JetBrains documents third-party and local models as working in AI Chat, while Junie and next-edit suggestions still require JetBrains’ cloud models. Expect chat to switch over and inline suggestions to stay on the built-in provider.
Use Test Connection in the provider settings to confirm the endpoint answers before assigning models to features.

​
Why this beats a bare tunnel

Without something in front of it, a local model server only listens on localhost or your LAN, and JetBrains’ cloud-hosted AI Assistant backend can’t reach either. A raw tunnel (for example ngrok) can expose the port, but it typically leaves the model listening with no authentication behind a URL that changes on every restart. Tokios’s connector is outbound-only, so your model never opens an inbound port, and every request to it carries a scoped sk-tok-… key:
  • No unauthenticated exposure — the connector dials out to Tokios; it never listens on a port, and requests need a valid key.
  • Stable URL — https://api.tokios.com/v1 doesn’t change between restarts, so you set it in AI Assistant once.
  • Scoped keys — a key can be limited to specific deployments and revoked instantly if it leaks, without touching your local model or connector.

​
Troubleshooting

The API key AI Assistant sent is missing, malformed, or has been revoked. Confirm you pasted the full sk-tok-… key with no extra whitespace, and check its status on the Keys tab.
The model value AI Assistant sent doesn’t match a registered deployment. Confirm the model name you entered in AI Assistant exactly matches the public name on the Models tab — not your Ollama/llama.cpp/vLLM/LM Studio local model id.
The connector for that deployment is offline or at its concurrency cap. Check the Connectors tab — the connector needs to show as Online, and the machine running it needs to still be up and reachable.

​
Next steps

Quickstart

Set up the connector, register a model, and create a key from scratch.

Compatible clients

See base URL and key setup for other OpenAI- and Anthropic-compatible clients.